Charge Dynamics in Organic Photovoltaic Materials: Interplay

Jun 2, 2015 - This paper discusses the mechanism of generation of free charges in organic photovoltaic cells (OPV) from electrostatically bound electr...
1 downloads 11 Views 1MB Size
Article pubs.acs.org/JPCC

Charge Dynamics in Organic Photovoltaic Materials: Interplay between Quantum Diffusion and Quantum Relaxation Myeong H. Lee,* Juan Aragó, and Alessandro Troisi Department of Chemistry and Centre for Scientific Computing, University of Warwick, Coventry CV4 7AL, U.K. ABSTRACT: This paper discusses the mechanism of generation of free charges in organic photovoltaic cells (OPV) from electrostatically bound electron−hole pairs. The efficiency of this process is explained when interfacial chargetransfer (CT) states are generated by direct optical excitation. We used semiclassical quantum dynamics at a short time scale (∼100 fs) and Redfield theory at a relatively long time scale (∼10−100 ps) to cover both the process of dissociation and the relaxation to the lowest energy state. Our calculations suggest that a CT state with an intermediate electron−hole separation can evolve into a charge-separated (CS) state on ultrafast time scales (∼100 fs) as a result of quantum diffusion. On long time scales, however, the CS states ultimately relax to the low-energy CT states due to the interaction with the thermal bath, indicating that the yield of free charge carrier generation is determined by the interplay between ultrafast charge separation, due to quantum diffusion, and the much slower quantum relaxation process. cells.13−22 For example, Tamura and Burghardt13 attributed fast CS to the two main factors, (i) electron delocalization within the fullerene condensate and (ii) the role of vibronically hot CT states, by carrying out a combined approach of electronic structure calculations and quantum dynamical analysis. Sun and Stafström14 showed, using nonadiabatic Ehrenfest dynamics, that intermolecular interactions strong enough to delocalize the wavefunctions on the acceptor side are crucial to overcome the Coulomb attraction. Similar conclusions have been obtained with a very elementary treatment of the electron−phonon coupling in ref 23. Smith and Chin15 suggested that delocalized electronic eigenstates of the acceptor crystallite can account for both ultrafast CS that occurs on short time scales after exciton dissociation and the separation of the relaxed CT state formed on long time scales before nongeminate recombination takes place. Vázquez and Troisi16 suggested that Frenkel excitons on donor material dissociate into partially charge-separated states, i.e., hot CT states in nature, which can explain efficient charge generation in OPV materials. All of these studies describe the formation of delocalized, higher energy CT states from the Frenkel excitonic states, which are easily transformed into free charges without energy penalty. In other words, it is the resonant coupling of photogenerated singlet excitons to a highenergy manifold of fullerene electronic states that enables efficient charge generation, bypassing localized CT states. However, it was recently shown experimentally11 that CS can occur very efficiently via an optically generated CT state. The theoretical studies so far, identifying a CS mechanism via the

1. INTRODUCTION Light absorption in donor/acceptor (D/A) heterojunction organic photovoltaic (OPV) materials leads to the formation of spatially localized electron−hole pairs (Frenkel excitons) rather than creating free charge carriers as in inorganic PV materials.1−3 Photocurrent generation in OPV devices, therefore, requires first the dissociation of Coulombically bound Frenkel excitons into interfacial charge transfer (CT) state excitons (D+A−) and then the separation of electron−hole pairs into free electron and hole carriers (D+ + A−). Because of the low dielectric constant, the binding energy of CT excitons in organic materials is typically an order of magnitude larger than the thermal energy. However, experimental evidence indicates unequivocally that the charge generation process occurs on the ultrafast time scale within a few hundred femtoseconds in OPV cells.4−8 Understanding how the tightly bound electron−hole pair overcomes the Coulomb barrier and dissociates into free charge carriers is therefore one of the long-standing questions in OPV and is crucial for the design of new devices with improved efficiency. Recent experiments using time-resolved two-photon photoemission (TR-2PPE) spectroscopy and transient absorption spectroscopy suggest that efficient CS is assisted by the hot charge-transfer (CT) state with excess photon energy.4,6,9 However, other experimental evidence indicates that the internal quantum efficiency is independent of whether or not initially generated excited states have higher energy than the thermally relaxed CT states.10,11 Therefore, the relevant question about what drives ultrafast CS in OPV cells is still open.12 A good number of theoretical studies have been performed to provide a quantitative description of the ultrafast CS in OPV © XXXX American Chemical Society

Received: April 27, 2015 Revised: June 1, 2015

A

DOI: 10.1021/acs.jpcc.5b03989 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C direct formation of higher energy and more delocalized CT states, are therefore unable to explain this new evidence. In ref 11, the lowest energy emissive interfacial CT state was directly and exclusively generated from the ground state via optical excitation, and the internal quantum efficiency (IQE, or the fraction of photons transformed into charges that are collected at the electrodes) was essentially similar to the IQE measured for incident radiation populating the exciton on the donor material. Direct photoexcitation of the thermally relaxed CT states has been previously carried out by other experimental groups,10,24,25 and the conclusion was, similarly, that the CT states with excess energy are not required for the free charge carrier generation. Herein we consider a scenario where the CT states are generated by direct optical excitation with an intermediate electron−hole separation and energy lower than the energy of free charges. Recently, Barker et al.26 showed that such CT states with the intermediate-range electron−hole separation exist by measuring distance distributions of photogenerated charge pairs for the wide range of OPV blends. They estimated that electron−hole separation in the range of 3−4 nm is critical for the CT states to yield free charge carriers, depending on the tunneling attenuation factor β. The possibility of generating charges in a range of distances near the interface is easily explained by theoretical modeling.27 CT states with a small hole−electron separation have larger oscillator strength but smaller density, while CT states at intermediate hole−electron separation are more easily formed because their higher density more than compensates for their weaker oscillator strengths. In this work we aim at finding out whether CT states at intermediate hole−electron separation can still separate into free holes and electrons as a result of quantum diffusion in the short time scale (∼100 fs). Even if charges can be separated on short time scales, the relaxation process of the CS states to lowlying CT states is expected to take place at relatively long time scales (∼10−100 ps). The same relaxation mechanism can be thought to play a role when free hole and electron are trapped in a CT state, which can be considered as an intermediate state for nongeminate charge recombination. To obtain a complete description of charge dynamics it is therefore important to model the charge dynamics of both processes, “low-energy CT to CS states” and “CS to low-energy CT states”, which would occur on two different time scales. To this end, we build a model Hamiltonian of the D/A interface and employ two different quantum dynamics methods (Ehrenfest dynamics and Redfield theory) that are expected to properly describe charge dynamics for each regime. We suggest that the competition between the ultrafast charge separation and comparatively slower relaxation determines ultimately the yield of free charge carrier generation.

Figure 1. Top panel: Schematic of a one-dimensional acceptor lattice model. Central panel: corresponding on-site energy εi (L = 13 Å and ϵ/ ϵ0 = 3.5). The arrows indicate the short time quantum diffusion and the long time relaxation toward the lowest energy site. Bottom panel: Schematics of the time evolution of charge density along the acceptor sites. el Ĥ =

N

∑ [εi|i⟩⟨i| + ∑ i=1

j=i±1

τ |i⟩⟨j|] (1)

where |i⟩ is an electronic state localized on molecule (site) i; N is the total number of sites; L is the distance between adjacent sites (this is also assumed to be the distance between the hole and electron nearest to the interface); εi = −(e2/(4πϵLi)) is the on-site energy determined by Coulomb interaction of an electron with a hole separated by distance L × i in a dielectric medium with a dielectric constant ϵ/ϵ0; and τ is the electronic coupling between adjacent states that is assumed to be the same for all adjacent pairs. Realistic charge dynamics is influenced by the electron− phonon coupling at short time and the coupling with the thermal bath at longer time, as discussed in the following sections. Nevertheless, it is also useful to consider the charge dynamics within the simplified Hamiltonian presented above to determine the features of quantum diffusion in the absence of coupling with other degrees of freedom. Charge dynamics governed by eq 1 can be described in two different ways. One is to obtain the time evolution of the wavefunction |ψ(t)⟩ = ∑iCi(t)|i⟩ by solving the time-dependent Schrödinger equation iℏ(∂/∂t)|ψ(t)⟩ = Ĥ el|ψ(t)⟩, and the other is to obtain time evolution of the density matrix ρ̂(t) by solving the quantum Liouville equation (∂/∂t)ρ̂(t) = −(i/ℏ)[Ĥ el, ρ̂(t)]. Population of charges on each site can be obtained by computing |Ci(t)|2 and ρii(t), respectively, which produces identical results in a closed system. In what follows, we set the electronic Hamiltonian parameters to N = 20, L = 13 Å, ϵ/ϵ0 = 3.5, and τ = 0.08 eV unless otherwise noted. (L = 13 Å is used to model fullerene derivatives as an acceptor molecule.) The permittivity of a dielectric medium ϵ is close to the reported experimental value,29,30 and τ is an intermediate value encountered in computational studies.7,14,15,23,27,31,32 Charge is initially local-

2. METHODS 2.1. Purely Electronic Dynamics. To build a model Hamiltonian we treat the acceptor crystal as a finite onedimensional lattice, where each acceptor molecule occupies each lattice point, while a hole on the donor molecule is treated as a point charge fixed in space (see Figure 1). A similar type of model Hamiltonian has been widely used to describe Frenkel exciton dissociation and charge dynamics in D/A heterojunctions.7,14−17,22,23,27,28 By considering a single electronic state per acceptor site and only the nearest-neighbor interaction between these electronic states the electronic Hamiltonian can be written as B

DOI: 10.1021/acs.jpcc.5b03989 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C

reference value of α(2) is set to 0.34 eV/Å, which was determined from the relationship σ τ = α (2) (2k B T/ m(2)(ω(2))2)1/2, where στ is the standard deviation of τ distribution and approximated to στ ≈ τ/333 for τ = 0.08 eV. Table 1 summarizes the parameters used in this work. In the

ized on the second nearest site to the interface (|ψ(t = 0)⟩ = |2⟩), which corresponds to the electron−hole separation of 2.6 nm. 2.2. Ehrenfest Dynamics. Thermal fluctuation of nuclei from their equilibrium position leads to modulation of the electronic energy levels and their couplings. In the presence of electron−phonon coupling, the total Hamiltonian can be written as tot el el − ph ph Ĥ = Ĥ + Ĥ + Ĥ

Table 1. Parameter Values Used in the Model Hamiltonian unless Otherwise Noteda Ĥ el

where el − ph Ĥ =

N

∑ ∑ [λ(k)ui(k)|i⟩⟨i| + ∑ k

i=1

α(k)(ui(k) − u(j k))|i⟩⟨j|] Ĥ el−ph

j=i±1

(3) ph Ĥ =

N

⎡ 1 (k) (k) 2 ⎤ 1 m (ui̇ ) + m(k)(ω(k)ui(k))2 ⎥ ⎦ 2 2

∑ ∑ ⎢⎣ k

i=1

(4)

and Ĥ el is given by eq 1. Here, {λ(k)} and {α(k)} are the local (or Holstein) and nonlocal (or Peierls) electron−phonon coupling constants, respectively, and {u(k) i } is the displacement of an atom on site i of the nuclear mode k. The first and the second term on the right-hand side of eq 3 represents the modulation of the on-site energy εi due to the displacement of the nuclei from their equilibrium position and the modulation of the intermolecular electronic coupling τ due to the displacement of the adjacent nuclei relative to each other, respectively. Within an Ehrenfest scheme the phonons are treated as classical harmonic oscillators as described in eq 4. Then, the equation of motion for the nuclei position can be written as m(k)uï (k) = −m(k)(ω(k))2 ui(k) − | ψ (t )⟩

value

description

L (Å) N ϵ/ϵ0 τ (eV) m(1) (amu) ω(1) (cm−1)

13 (5) 20 (10, 30, 50) 3.5 0.08 (0.04, 0.16) 6 1500

Λ (eV) λ (eV/Å)

(0.01−0.4) (0.71−4.48)

m(2) (amu) ω(2) (cm−1)

910 (392) 50

α (eV/Å)

0.34 (0.03−1.36)

γ−1 (fs)

50 (100)

lattice spacing number of acceptor sites dielectric constant electronic coupling mass of phonon mode (1) frequency of phonon mode (1) reorganization energy local electron−phonon coupling mass of phonon mode (2) frequency of phonon mode (2) nonlocal electron−phonon coupling characteristic time scale of phonon spectral density

(2)

∂ ∂ui(k)

Ĥ ph

a Parameter range used to see its effect on charge dynamics is listed in parentheses (m(2) = 392 amu, which corresponds to the mass of a perylene-3,4,9,10-tetracarboxylic dianhydride (PTCDA), is used when L = 5 Å).

following we investigate the effect of τ, λ, and α on charge dynamics. (We omit the superscripts in λ(k) and α(k) since we consider only one effective mode for each type of electron− phonon coupling.) 2.3. Reduced Density Matrix Dynamics. The relaxation process is expected to dominate charge dynamics on long time scales due to the interaction with the thermal bath, resulting in the relaxation of the CS states to the lower-lying CT states. A major drawback of Ehrenfest dynamics is that it fails to achieve thermal equilibrium,39,40 e.g., non-Boltzmann distribution of equilibrium state population, and does not properly account for decoherence.41−43 Several approaches are being debated in the literature to improve long-time behavior44−46 and incorporate decoherence effects41,47 in Ehrenfest dynamics, but they are outside the scope of our analysis. An alternative way to study the system relaxation process is to employ the density matrix approach where relaxation dynamics is described in terms of the reduced density operator

el el − ph ) ⟨ψ (t )|(Ĥ + Ĥ

(5)

The last term in eq 5 represents the effect of the electronic Hamiltonian coupled with the phonons on the nuclear degrees of freedom. Here, initial positions, {u(k) i (0)}, and velocities, {u(k) i (0)}, of the nuclei are selected from the Boltzmann distribution at room temperature. Details of numerical integration are provided in ref 33. The model parameters can be obtained by quantum chemistry calculations and classical molecular dynamics simulations for realistic molecular systems. As in this work we are trying to establish the basic physical principles, we will set the parameters to previously determined literature values and study their effect on the results. The Holstein coupling constant λ is associated with the reorganization energy Λ. Employing a single effective high frequency mode (k = 1) for λ is proven to be valid for organic chromophores in charge transfer reactions.34 In this case, λ(k) can be written as λ(1) = ω(1)(m(1)Λ)1/2 and λ(k) = 0 for k ≠ 1. We set ω(1) = 1500 cm−1 and m(1) = 6 amu, which corresponds to the C−C/CC stretching normal mode throughout the simulations.35 In general, low frequency modes are responsible for modulating the intermolecular electronic coupling.36 We assume that this modulation can be represented by a single effective low frequency mode ω(2) on the basis of previous computational work.35,37 Here, we set ω(2) = 50 cm−1, which was obtained from the spectral density of τ for the organic system,35,38 and m(2) = 910 amu, which corresponds to the total mass of phenyl-C61-butyric acid methyl ester (PCBM). The

̂ }= ρ ̂ ≡ TrB{ρ tot

̂ |b⟩ ∑ ⟨b|ρ tot b

(6)

where ρ̂tot is the density operator of the “system + bath”; TrB{} represents the partial trace over the bath degrees of freedom; and {|b⟩} denotes the eigenstates of the bath Hamiltonian. We assume that the system−bath coupling is weak and the bath dynamics is much faster than the system dynamics, and therefore memory effects in the reduced density dynamics can be ignored (Markov approximation). It should be noted that Markov approximation is not valid at short times (∼fs) when the characteristic time scale of the system is faster than the time scales on which the bath correlation function decays. On these time scales, however, we do not expect any significant C

DOI: 10.1021/acs.jpcc.5b03989 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C

Figure 2. Time evolution of site population in the absence of electron−phonon coupling. On the right panel color-mapped population (white to red when going from ρii = 0 to ρii = 1) is plotted where the vertical axis denotes a site index. Charge is initially localized on the site i = 2 in the upper panel and i = 1 in the lower panel, respectively.

relaxation/dephasing to take place. As we will see, the CS process with reasonable system parameters cannot be possibly Markovian, while the relaxation process probably satisfies the Markovian limit. The transition between the two regimes is in general complicated and will not be studied in detail in this work. We, however, discuss the possible interpolation between the two regimes in the Results section. In the Markovian limit the equation of motion for the reduced density operator can be described by the Redfield equation48−50 ∂ ρ (t ) = −iωμνρμν (t ) + ∂t μν

where the ℭij [ω] is the Fourier transform of the ℭij (t ), ∞

ℭij [ω] ≡ ∫ dteiωt ℭij (t ), and expressed in terms of bath 0

spectral density Jij(ω). The imaginary parts of ℭ[ω] represent bath-induced shift of the transition frequency and do not lead to relaxation. Therefore, we only consider the real parts of ℭ[ω] that are relevant to the system relaxation and given by51 Re ℭij [ω] = ℏJij (ω)[nBE(ω) + 1]

where nBE(ω) ≡ [eβℏω − 1]−1 is the Bose−Einstein distribution function. In this work we employ Drude−Lorentz bath spectral density J(ω) = 2Λ(ωγ/(ω2 + γ2))53,54 and assume Jij(ω) = δijJ(ω).51 Here, the damping constant γ determines the width and peak position of the spectral density (Λ is the reorganization energy as defined in section 2.2). It should be noted that while only a few local phonon modes are selected when employing Ehrenfest dynamics to describe CS dynamics the continuous spectrum of phonon modes, J(ω), is employed when using Redfield theory, and only the local electron− phonon coupling is considered. Once ρμν is obtained from eq 7, it can be transformed to the site representation by ρ̂site = Uρ̂̂Û −1, where Û is the unitary operator that diagonalizes Ĥ el via Û −1Ĥ elÛ . Despite the restrictive assumptions described above, Redfield theory has been shown to provide a simple, computationally efficient, and valuable way to study dissipation due to a thermal environment for a variety of problems.52,55−58 Recently, Kelly et al.59 proposed a method that combines Ehrenfest mean field theory with the reduced density matrix formalism, which is nonperturbative, non-Markovian, and nonrestrictive on the form of the Hamiltonian with a much lower computational cost than mean-field theory. In this work we choose two different approaches, Ehrenfest dynamics in short times and Redfield theory in long times, which are known to be well-suited in each regime.

∑ Rμν ,μ′ ν′ρμ′ ν′(t ) (7)

μ′ν′

where μ, ν stand for eigenstates of the system Hamiltonian and ωμν ≡ (Eμ − Eν)/ℏ, where Eμ, Eν are the corresponding eigenvalues. The first term on the right-hand side of eq 7 describes system dynamics in the absence of dissipation. System relaxation due to the interaction with the thermal bath is described by the second term via the Redfield tensor Rμν,μ′ν′, which can be expressed in terms of the damping matrix Rμν , μ ′ ν ′ ≡ Γν ′ ν , μμ ′ + Γ*μ ′ μ , νν ′ − δνν ′ ∑ Γμκ , κμ ′ κ

− δμμ ′ ∑ Γ*νκ , κν ′ κ

(8)

where the damping matrix Γμν,μ′ν′ is given by Fourier−Laplace transforms of the coupling correlation function50 and mainly determined by the bath correlation functions ℭ(t ).49,50 For the system−bath interaction that is linear in the displacement of (k) the bath oscillators, Ĥ el−ph = ∑k∑Ni=1λ(k) i ui |i⟩⟨i|, the damping 51,52 matrix can be written as Γμν , μ ′ ν ′ ≡

1 ℏ2

∑ ⟨eμ|i⟩⟨i|eν⟩⟨eμ′|j⟩⟨j|eν′⟩ℭij[ων′ μ′] i,j

(10)

(9) D

DOI: 10.1021/acs.jpcc.5b03989 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C

Figure 3. (a) Expectation value of position ⟨x⟩outgoing/L (L = 13 Å) for N = 50 in the absence of electron−phonon coupling (α = λ = 0). (b) ⟨x⟩outgoing/L, averaged over 300 trajectories, for different reorganization energies Λ with α = 0, τ = 0.08 eV, and N = 20. (c) ⟨x⟩outgoing/L, averaged over 300 trajectories, for different nonlocal electron−phonon coupling constants α with λ = 0, τ = 0.08 eV, and N = 20. (d) Fraction of outgoing charge η for different λ and α with τ = 0.08 eV and N = 20. Each line corresponds to different values of Λ (varying from 0 to 0.4 eV as in (b)).

3. RESULTS AND DISCUSSION 3.1. Charge Dynamics in the Absence of Electron− Phonon Coupling. It is instructive to analyze charge dynamics in the absence of electron−phonon coupling. In fact, while the latter has been proven essential for the quantitative description of exciton dissociation,18,21,60 many qualitative features are present in the simplest possible model without electron−phonon coupling,23 and it is convenient to isolate the phenomenology of a simple, purely electronic Hamiltonian. Figure 2 (upper panel) shows the time evolution of population ρii(t) of several sites (i = 1−5, 20) when charge is localized on the site i = 2 at t = 0. Here, we analyze charge dynamics until the outgoing wavepacket reaches the boundary (set by N = 20). Charge propagates with time along the acceptor sites, leaving the D/A interface, and reaches i = 20, which corresponds to the electron−hole separation of 26 nm, after 90 fs with population ∼0.35. Note that the population at the lowest energy CT state, ρ11, does not decay to zero. Instead, it fluctuates around 0.1−0.2 with time. This bifurcation behavior of the wavefunction can be more easily seen by plotting the population density map (upper-right panel of Figure 2), where population of each site is color-mapped (red for ρ̂ii = 1 and white for ρ̂ii = 0) in the two-dimensional plot (site index on the vertical axis and time on the horizontal axis). It illustrates that the charge population follows two different paths: one propagates outward from the D/A interface, and the other is relaxed to the lowest energy CT state. We note in passing that the outgoing wavepacket spreads over a few sites and shows slight increase in the spread as time goes on, whereas the charge density relaxed to the lowest energy CT state is more or less localized on i = 1. On the other hand, when the initial state is localized on the lowest energy CT state (i = 1), the propagation of the wavefunction away from the interface becomes much weaker. In the lower panel of Figure 2 site population exhibits sign of charge propagation in early times; however, it fades away with

time (ρii decreases fast as i increases), while the population at the lowest energy CT state remains high (ρ11 = 0.5−0.7). Significantly weakened charge propagation behavior is also illustrated by the population density map plotted in the lowerright panel of Figure 2, where the majority of the charge population remains at the lowest energy CT state (i = 1), and population at the sites far from the interface is almost negligible. This result indicates that the CS process is not likely to be initiated by a charge located at the shortest possible distance from the interface, but as long as the initial state possesses an intermediate-range electron−hole separation (e.g., a CT state with 2.6 nm electron−hole separation), the generation of free charge due to quantum diffusion becomes possible. Electron−hole separation of 2.6 nm that is required for efficient CS to occur is close to the lower bound of the critical distance of the charge pairs (3−4 nm) to generate free charges found in ref 26. The expectation value of the energy is constant and below the energy of free charges. However, as the system evolves quantum mechanically, it will have finite probability to be found at higher (or lower) energy sites. In this case, the fast propagation toward higher energy states is facilitated by their higher density. This quantum diffusion process (which as we will see in the next section is slowed down by electron−phonon coupling) cannot take place indefinitely, and at some point, the interaction with the thermal bath will induce dephasing and thermal relaxation toward the lowest energy state as discussed in section 2.3. Before we proceed to analyze the effect of the electronic Hamiltonian parameters on charge dynamics, we define a few physical quantities for the quantitative analysis of charge dynamics. One simple way to analyze charge dynamics is to obtain the expectation value of the position ⟨x⟩ as a function of time. By assuming ⟨i|x|i′⟩ = xiδi,i′ = iLδi,i′, ⟨x⟩ can be simply obtained by E

DOI: 10.1021/acs.jpcc.5b03989 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C N

⟨x⟩ =

∑ iLρii (t )

(11)

i=1

However, this definition can give misleading information if the distribution of ρii(t) is multimodal, as seen in Figure 2. Therefore, to monitor the propagation of wavepacket leaving the D/A interface, we additionally define the expectation value of the position of the outgoing wavefunction as follows ⟨x⟩outgoing ≡

∑i ≥ i iLρii (t ) 0

∑i ≥ i ρii (t ) 0

(12)

where i0 denotes the site where the charge is initially localized. Note that i0 is included in the summation of eq 12, so that ⟨x⟩outgoing at t = 0 gives the initial charge position. The expectation value of the position provides an estimate of electron−hole pair distance as a function of time, but it does not provide the information on what fraction of charge is leaving the interface. For instance, a small charge at large distances is equivalent to a large charge at small distances. Therefore, we obtain the fraction of outgoing charge η defined as η≡

∑ ρii (t ) i > i0

(13)

which also provides the probability of the wavefunction to be found in the higher-lying states than the initial state. Figure 3a shows the ⟨x⟩outgoing as a function of time for different intermolecular electronic couplings τ. To avoid the boundary reflection during charge propagation, arising from the finite size of the acceptor lattice model when τ is large, we increase the number of lattice points to N = 50 in this case. As expected, the charge leaves the interface at shorter times as τ increases. For instance, it takes 12 fs to reach the site i = 5 when τ = 0.16 eV, whereas it takes 20 fs when τ = 0.08 eV and 70 fs when τ = 0.04 eV. When τ is very small (= 0.04 eV), the CS process becomes very inefficient. It occurs much more slowly and almost stops after 40 fs. Our result that sufficient intermolecular electronic coupling is necessary for CS to occur is in line with other studies14,15,32 where the increase of intermolecular interactions leads to more delocalization of the acceptor sites and therefore facilitates CS. We also analyzed charge dynamics for L = 5 Å to model flat acceptor molecules (results not shown). As expected, CS dynamics is similar regardless of L if the charge is initially localized with similar electron−hole separation (e.g., we find similar CS dynamics when charge is localized on the site i = 4 for L = 5 Å and when localized on i = 2 for L = 13 Å). 3.2. Charge Dynamics on Short Time Scales Using the Ehrenfest Scheme. Now we introduce the electron−phonon coupling in the model Hamiltonian to see its effect on charge dynamics. First, we investigate the effect of local electron− phonon coupling λ on charge dynamics. To this end, we set nonlocal electron−phonon coupling α to zero and vary only reorganization energy Λ, which determines electron−phonon coupling λ by the relationship λ = ω(mΛ)1/2. The charge population density map obtained for different Λ (Figure 4) shows that the charge propagation feature becomes weaker as Λ increases, and as a result the fraction of charge density at site i = 1 (ρ11) increases with Λ. Slowing down of the CS process can be quantitatively analyzed by plotting ⟨x⟩outgoing/L as shown in Figure 3b. For instance, as Λ increases from 0 to 0.4 eV,

Figure 4. Population density map for Λ = 0.05 eV (top), 0.1 eV (middle), and 0.2 eV (bottom). Charge is initially localized on the site i = 2, and α is set to zero in all cases.

⟨x⟩outgoing/L at t = 85 fs decreases from 16 to 5.4, which corresponds to an electron−hole separation of 21 and 7 nm, respectively. Next, we increase α from 0.03 to 1.36 eV/Å while keeping λ = 0. We find a similar trend of ⟨x⟩outgoing/L in Figure 3c that the CS process slows down as α increases. Note that as α increases to 0.68 eV/Å or above, CS becomes so slow that increasing α hardly affects charge dynamics any more. We now vary both λ and α, and plot η at t = 80 fs in Figure 3d. Note that η is as high as 0.7−0.8 when both Λ and α remain relatively small. As electron−phonon coupling further increases, η decreases. However, we find η ≥ 0.4 even for the largest electron−phonon coupling constants considered here. Considering the range of reorganization energies (Λ = 0.05−0.3 eV)31,61−64 and the distribution of intermolecular electron− phonon couplings33,35 in organic semiconducting materials, we can conclude that the CS process can be quite efficient in typical OPV materials. For instance, for the system parameters typical for organic materials (Λ = 0.2 eV, α = 0.34 eV/Å) we find η = 0.51. This represents one of our main results, suggesting that CS can be achieved with a ∼50% probability in the presence of realistic electron−phonon couplings, if the F

DOI: 10.1021/acs.jpcc.5b03989 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C

Figure 5. (a) Population of the five sites near the D/A interface. (b) Expectation value of the position ⟨x⟩/L for different values of Λ with γ−1 = 50 fs. Dashed curve is for γ−1 = 100 fs and Λ = 0.02 eV. Shaded region corresponds to the regime where the charge dynamics cannot be accurately described by the employed quantum dynamics scheme.

Figure 5b also shows how parameters Λ and γ affect charge dynamics. As reorganization energy Λ increases, charge propagation occurs more slowly, and therefore the electron is separated from the hole by a shorter distance as seen in the previous section. The time scale on which charge relaxation occurs, on the other hand, is only slightly affected, and the overall behavior of charge dynamics does not critically change with the reorganization energy. We point out that eq 7 holds only in the limit of weak system−bath coupling, and therefore we keep the value of Λ smaller than τ (= 0.08 eV). The black dashed curve in Figure 5b shows the ⟨x⟩/L for γ−1 = 100 fs, which is the typical value in photosynthetic excitation energy transfer processes.51,54,66 When compared with the black solid curve with γ−1 = 50 fs, system relaxation occurs on a slightly longer time scale as the damping constant γ decreases, as expected. On the other hand, charge dynamics on short time scales is hardly affected by γ because the system dynamics in this regime is mostly dominated by the quantum diffusion process rather than quantum relaxation. We found a similar trend of charge dynamics except for a slight increase in the maximal value of ⟨x⟩ by 1−2 nm when varying γ−1 from 50 to 100 fs for larger values of Λ (=0.04 and 0.06 eV). Therefore, we conclude that although the distance of separated charge pairs can vary within a few nanometers depending on the system− bath or bath parameters the overall behavior of charge dynamics does not significantly change, and charge relaxation occurs on a time scale of ∼10 ps. We additionally performed quantum dynamics calculation for different system parameters, such as varying the D−A distance from 9 to 19 Å while keeping lattice spacing the same (L = 13 Å), increasing the number of sites from N = 10 to N = 30 and extending the acceptor model from one-dimensional lattice to two-dimensional lattice for a different number of sites along the interface. In all cases, we found a similar trend of charge dynamics that charges can separate on the ultrafast time scale on the order of tens of femtoseconds due to quantum diffusion, and they eventually relax to the lowest-energy CT state on a longer time scale on the order of 1−10 ps due to the interaction with the thermal bath. In our model the only possible fate for the electron carrier is to relax to a bound CT state, from which hole−electron annihilationand therefore efficiency lossis the most likely outcome. In a real open system, within the few picoseconds of relatively large hole−electron separation, the hole can diffuse away from the interface, and the electron can diffuse to a distance where it becomes equilibrated with the quasi-Fermi level of the electron acceptor. The picosecond time scale is the

initial CT state can be generated with an intermediate-range electron−hole separation. 3.3. Charge Dynamics on Long Time Scales Using the Redfield Equation. We now present charge dynamics on long time scales obtained by the Redfield equation. It should be noted that short time dynamics is also correct in this case because the effect of the bath is insignificant at very short times. Figure 5a shows the population (diagonal elements of ρ̂site) for the five sites near the D/A interface using the parameters γ−1 = 50 fs and Λ = 0.05 eV. There is comparatively less information on γ, and therefore we try a plausible range of γ values as discussed below. We expect that system relaxation occurs much more slowly than the correlation time of the bath fluctuation, γ−1 = 50 fs, and therefore the Markov approximation is valid in this regime. The time evolution of the site population exhibits outgoing charge propagation on short time scales (less than 100 fs), as seen in the previous section using the Ehrenfest method. However, after CS takes place on a short time scale, lower-lying CT states begin to gain population as the quantum relaxation process dominates system dynamics. Site population eventually reaches a plateau after ∼4 ps where ρ11 and ρ22 are largest and the population of the higher-lying CT states is negligible. Two different processes governing charge dynamics on two different time scales (quantum diffusion on short time scales and quantum relaxation on long time scales) can be easily seen by examining the expectation value of the position. Since we expect charge to relax eventually to the lowest-lying CT states, we evaluate the expectation value of position ⟨x⟩ instead of evaluating the position of outgoing charge ⟨x⟩outgoing. Figure 5b shows ⟨x⟩/L for different reorganization energies Λ. It starts from ⟨x⟩/L = 2, where charge is initially populated, increases until it reaches the largest electron−hole separation on a time scale of ∼100 fs, and then decreases until it reaches an equilibrium value (⟨x⟩EQ/L = 1.3) on a time scale of ∼10 ps.65 It should be noted that Markov approximation employed here is valid when the time scale of bath correlation function is faster than that of system dynamics. Since the relaxation process will not take effect at very short times, the shaded region in Figure 5b approximately represents the intermediate regime where the quantum dynamics scheme employed in this work may not be valid. We speculate, however, that a qualitative picture of charge dynamics in this regime would not significantly deviate from what is described here except for possible variations in CS distances and time scales, based on charge dynamics on long time scales. We note in passing that Ehrenfest dynamics does not reach correct equilibrium as described in section 2.3. For instance, we obtain ρ11 = 0.1 and ⟨x⟩/L ∼ 10 at equilibrium. G

DOI: 10.1021/acs.jpcc.5b03989 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C Notes

one appropriate for device modeling via kinetic Monte Carlo techniques,67−69 which take into consideration the charge extraction from the electrodes. Conversely, the ultrafast time scale is not suitably described by kinetic methodologies because the charge diffusion takes place coherently for the first hundreds of femtoseconds. Our model accounts for both coherent charge dynamics occurring on a time scale of hundreds of femtoseconds that can be detected by ultrafast spectroscopy measurements and charge relaxation to an equilibrium where standard device modeling techniques (kinetic Monte Carlo, drift-diffusion equations) can be applied to predict microscopic properties of OPV devices. Therefore, the phenomenology described here offers a possible link between the observations in ultrafast pump−probe laser experiments and the electrical measurements of the devices. Finally, one can note that the methodology used to describe the process of relaxation to a bound CT state from a quasi-CS state can be also used to describe the formation of a bound CT state from nongeminate hole−electron pairs.

The authors declare no competing financial interest.



ACKNOWLEDGMENTS This work was supported by ERC (Grant No. 615834) and a Marie Curie Intra European Fellowship (FP7-PEOPLE-2012IEF-329513).



(1) Kippelen, B.; Brédas, J.-L. Organic Photovoltaics. Energy Environ. Sci. 2009, 2, 251−261. (2) Benanti, T. L.; Venkataraman, D. Organic Solar Cells: An Overview Focusing on Active Layer Morphology. Photosynth. Res. 2006, 87, 73−81. (3) Gregg, B. A.; Hanna, M. C. Comparing Organic to Inorganic Photovoltaic Cells: Theory Experiment, and Simulation. J. Appl. Phys. 2003, 93, 3605−3614. (4) Muntwiler, M.; Yang, Q.; Tisdale, W. A.; Zhu, X.-Y. Coulomb Barrier for Charge Separation at an Organic Semiconductor Interface. Phys. Rev. Lett. 2008, 101, 196403. (5) Grancini, G.; Maiuri, M.; Fazzi, D.; Petrozza, A.; Egelhaaf, H.-J.; Brida, D.; Cerullo, G.; Lanzani, G. Hot Exciton Dissociation in Polymer Solar Cells. Nat. Mater. 2013, 12, 29−33. (6) Jailaubekov, A. E.; Willard, A. P.; Tritsch, J. R.; Chan, W.-L.; Sai, N.; Gearba, R.; Kaake, L. G.; Williams, K. J.; Leung, K.; Rossky, P. J.; et al. Hot Charge-Transfer Excitons Set the Time Limit for Charge Separation at Donor/Acceptor Interfaces in Organic Photovoltaics. Nat. Mater. 2013, 12, 66−73. (7) Gélinas, S.; Rao, A.; Kumar, A.; Smith, S. L.; Chin, A. W.; Clark, J.; van der Poll, T. S.; Bazan, G. C.; Friend, R. H. Ultrafast Long-Range Charge Separation in Organic Semiconductor Photovoltaic Diodes. Science 2014, 343, 512−516. (8) Provencher, F.; Bérubé, N.; Parker, A. W.; Greetham, G. M.; Towrie, M.; Hellmann, C.; Côté, M.; Stingelin, N.; Silva, C.; Hayes, S. C. Direct Observation of Ultrafast Long-Range Charge Separation at Polymer-Fullerene Heterojunctions. Nature Commun. 2014, 5, 4288. (9) Ohkita, H.; Cook, S.; Astuti, Y.; Duffy, W.; Tierney, S.; Zhang, W.; Heeney, M.; McCulloch, I.; Nelson, J.; Bradley, D. D. C.; et al. Charge Carrier Formation in Polythiophene/Fullerene Blend Films Studied by Transient Absorption Spectroscopy. J. Am. Chem. Soc. 2008, 130, 3030−3042. (10) Lee, J.; Vandewal, K.; Yost, S. R.; Bahlke, M. E.; Goris, L.; Baldo, M. A.; Manca, J. V.; Voorhis, T. V. Charge Transfer State versus Hot Exciton Dissociation in Polymer-Fullerene Blended Solar Cells. J. Am. Chem. Soc. 2010, 132, 11878−11880. (11) Vandewal, K.; Albrecht, S.; Hoke, E. T.; Graham, K. R.; Widmer, J.; Douglas, J. D.; Schubert, M.; Mateker, W. R.; Bloking, J. T.; Burkhard, G. F.; et al. Efficient Charge Generation by Relaxed ChargeTransfer States at Organic Interfaces. Nat. Mater. 2014, 13, 63−68. (12) Gao, F.; Inganäs, O. Charge generation in polymer-fullerene bulk-heterojunction solar cells. Phys. Chem. Chem. Phys. 2014, 16, 20291−20304. (13) Tamura, H.; Burghardt, I. Ultrafast Charge Separation in Organic Photovoltaics Enhanced by Charge Delocalization and Vibronically Hot Exciton Dissociation. J. Am. Chem. Soc. 2013, 135, 16364−16367. (14) Sun, Z.; Stafström, S. Dynamics of Charge Separation at an Organic Donor-Acceptor Interface. Phys. Rev. B 2014, 90, 115420. (15) Smith, S. L.; Chin, A. W. Ultrafast Charge Separation and Nongeminate Electron-Hole Recombination in Organic Photovoltaics. Phys. Chem. Chem. Phys. 2014, 16, 20305−20309. (16) Vázquez, H.; Troisi, A. Calculation of Rates of Exciton Dissociation into Hot Charge-Transfer States in Model Organic Photovoltaic Interfaces. Phys. Rev. B 2013, 88, 205304. (17) Caruso, D.; Troisi, A. Long-Range Exciton Dissociation in Organic Solar Cells. Proc. Natl. Acad. Sci. U.S.A. 2012, 109, 13498− 13502.

4. CONCLUSION In this paper we modeled charge dynamics in OPV materials for both short (∼fs) and relatively long time scales (∼100 ps) using a simple acceptor lattice model combined with the mixed quantum-classical/quantum dynamics methods in each regime. We have shown that CT states optically generated with an intermediate electron−hole distance can separate into CS states (free electrons and holes) on the ultrafast time scale without the need for the formation of hot/delocalized CT states, which can explain recent experimental findings that CT states with excess energy are not required for the free charge carrier generation in OPV materials. In the presence of both local and nonlocal electron−phonon couplings, charge separation is slowed down but still achieved with ∼50% probability. On long time scales, however, the quantum relaxation process eventually governs charge dynamics, and CS states relax to the lowest energy CT states, which can form a bound polaron pair with a hole and undergo the nongeminate charge recombination process. Our findings, therefore, suggest that the yield of free charge carrier generation in OPV devices may result from the interplay between the two competing processes: ultrafast charge separation resulting from quantum diffusion and much slower quantum relaxation arising from the interaction with the environment, thereby offering a new insight into the mechanism of free charge carrier generation in OPV materials. We have verified that the results are not qualitatively influenced by the choice of the system parameter. On the other hand, we have noted that a number of simplifications are necessary to achieve a description of the quantum relaxation in the model system. For example, the weak system−bath coupling limit might not be fully valid for OPV materials, and in any case, it is essential to predict the characteristic time scale for quantum relaxation from the material characteristics rather than phenomenological models. Employing quantum dynamics methods that are less restrictive and material specific appears to be essential to describe accurately the dynamics of charge generation.



REFERENCES

AUTHOR INFORMATION

Corresponding Author

*Phone: +44 24 7652 2261. E-mail: [email protected]. uk. H

DOI: 10.1021/acs.jpcc.5b03989 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C (18) Bittner, E. R.; Silva, C. Noise-Induced Quantum Coherence Drives Photo-Carrier Generation Dynamics at Polymeric Semiconductor Heterojunctions. Nature Commun. 2014, 5, 3119. (19) Duchemin, I.; Blase, X. Resonant Hot Charge-Transfer Excitations in Fullerene-Porpyrin Complexes: Many-Body BetheSalpeter Study. Phys. Rev. B 2013, 87, 245412. (20) Savoie, B. M.; Rao, A.; Bakulin, A. A.; Gelinas, S.; Movaghar, B.; Friend, R. H.; Marks, T. J.; Ratner, M. A. Unequal Partnership: Asymmetric Roles of Polymeric Donor and Fullerene Acceptor in Generating Free Charge. J. Am. Chem. Soc. 2014, 136, 2876−2884. (21) Han, L.; Zhong, X.; Liang, W.; Zhao, Y. Energy Relaxation and Separation of a Hot Electron-Hole Pair in Organic Aggregates from a Time-Dependent Wavepacket Diffusion Method. J. Chem. Phys. 2014, 140, 214107. (22) Kocherzhenko, A. A.; Lee, D.; Forsuelo, M. A.; Whaley, K. B. Coherent and Incoherent Contributions to Charge Separation in Multichromophore Systems. J. Phys. Chem. C 2015, 119, 7590−7603. (23) Troisi, A. How Quasi-Free Holes and Electrons Are Generated in Organic Photovoltaic Interfaces. Faraday Discuss. 2013, 163, 377− 392. (24) van der Hofstad, T. G. J.; Nuzzo, D. D.; van den Berg, M.; Janssen, R. A. J.; Meskers, S. C. J. Influence of Photon Excess Energy on Charge Carrier Dynamics in a Polymer-Fullerene Solar Cell. Adv. Energy Mater. 2012, 2, 1095−1099. (25) Parkinson, P.; Lloyd-Hughes, J.; Johnston, M. B.; Herz, L. M. Efficient Generation of Charges via Below-Gap Photoexcitation of Polymer-Fullerene Blend Films Investigated by Terahertz Spectroscopy. Phys. Rev. B 2008, 78, 115321. (26) Barker, A. J.; Chen, K.; Hodgkiss, J. M. Distance Distributions of Photogenerated Charge Pairs in Organic Photovoltaic Cells. J. Am. Chem. Soc. 2014, 136, 12018−12026. (27) Ma, H.; Troisi, A. Direct Optical Generation of Long-Range Charge Transfer States in Organic Photovoltaics. Adv. Mater. 2014, 26, 6163−6167. (28) Sun, Z.; Stafström, S. Dynamics of Exciton Dissociation in Donor-Acceptor Polymer Heterojunctions. J. Chem. Phys. 2013, 138, 164905. (29) Koster, L. J. A.; Mihailetchi, V. D.; Blom, P. W. M. Ultimate Efficiency of Polymer/Fullerene Bulk Heterojunction Solar Cells. Appl. Phys. Lett. 2006, 88, 093511. (30) Moulé, A. J.; Meerholz, K. Interference Method for the Determination of the Complex Refractive Index of Thin Polymer Layers. Appl. Phys. Lett. 2007, 91, 061901. (31) Coropceanu, V.; Cornil, J.; da Silva Filho, D. A.; Olivier, Y.; Silbey, R.; Brédas, J.-L. Charge Transport in Organic Semiconductors. Chem. Rev. 2007, 107, 926−952. (32) Tamura, H.; Tsukada, M. Role of Intermolecular Charge Delocalization on Electron Transport in Fullerene Aggregates. Phys. Rev. B 2012, 85, 054301. (33) Troisi, A.; Orlandi, G. Charge-Transport Regime of Crystalline Organic Semiconductors: Diffusion Limit by Thermal Off-Diagonal Electronic Disorder. Phys. Rev. Lett. 2006, 96, 086601. (34) Jortner, J. Temperature Dependent Activation Energy for Electron Transfer between Biological Molecules. J. Chem. Phys. 1976, 64, 4860. (35) Troisi, A. Predictions of the Absolute Charge Mobility of Molecular Semiconductors: the Case of Rubrene. Adv. Mater. 2007, 19, 2000−2004. (36) Troisi, A.; Orlandi, G. Dynamics of the Intermolecular Transfer Integral in Crystalline Organic Semiconductors. J. Phys. Chem. A 2006, 110, 4065−4070. (37) Troisi, A. Dynamic Disorder in Molecular Semiconductors: Charge Transport in Two Dimensions. J. Chem. Phys. 2011, 134, 034702. (38) Cheung, D. L.; Troisi, A. Theoretical Study of the Organic Photovoltaic Electron Acceptor PCBM: Morphology, Electronic Structure, and Charge Localization. J. Phys. Chem. C 2010, 114, 20479−20488.

(39) Parandekar, P. V.; Tully, J. C. Mixed Quantum-Classical Equilibrium. J. Chem. Phys. 2005, 122, 094102. (40) Parandekar, P. V.; Tully, J. C. Detailed Balance in Ehrenfest Mixed Quantum-Classical Dynamics. J. Chem. Theory Comput. 2006, 2, 229−235. (41) Akimov, A. V.; Long, R.; Prezhdo, O. V. Coherence Penalty Functional: A Simple Method for Adding Decoherence in Ehrenfest Dynamics. J. Chem. Phys. 2014, 140, 194107. (42) Yonehara, T.; Hanasaki, K.; Takatsuka, K. Fundamental Approaches to Nonadiabaticity: Toward a Chemical Theory beyond the Born-Oppenheimer Paradigm. Chem. Rev. 2012, 112, 499−542. (43) Käb, G. Statistical Mechanics of Mean Field Ehrenfest Quantum/Classical Molecular Dynamics: The Damped Harmonic Oscillator. J. Phys. Chem. A 2004, 108, 8866−8877. (44) Bastida, A.; Cruz, C.; Zúñiga, J.; Requena, A.; Miguel, B. A Modified Ehrenfest Method that Achieves Boltzmann Quantum State Populations. Chem. Phys. Lett. 2006, 417, 53−57. (45) Bastida, A.; Cruz, C.; Zúñiga, J.; Requena, A.; Miguel, B. The Ehrenfest Method with Quantum Corrections to Simulate the Relaxation of Molecules in Solution: Equilibrium and Dynamics. J. Chem. Phys. 2007, 126, 014503. (46) Aghtar, M.; Liebers, J.; Strümpfer, J.; Schulten, K. Juxtaposing Density Matrix and Classical Path-Based Wave Packet Dynamics. J. Chem. Phys. 2012, 136, 214101. (47) Subotnik, J. E. Augmented Ehrenfest Dynamics Yields a Rate for Surface Hopping. J. Chem. Phys. 2010, 132, 134112. (48) Redfield, A. G. On the Theory of Relaxation Processes. IBM J. Res. Dev. 1957, 1, 19. (49) May, V.; Kühn, O. Charge and Energy Transfer Dynamics in Molecular Systems, 3rd ed.; Willey-VCH: Berlin, 2011. (50) Nitzan, A. Chemical Dynamics in Condensed Phases; Oxford University Press: New York, 2006. (51) Ishizaki, A.; Fleming, G. R. On the Adequacy of the Redfield Equation and Related Approaches to the Study of Quantum Dynamics in Electronic Energy Transfer. J. Chem. Phys. 2009, 130, 234110. (52) Egorova, D.; Thoss, M.; Domcke, W.; Wang, H. Modeling of Ultrafast Electron-Transfer Processes: Validity of Multilevel Redfield Theory. J. Chem. Phys. 2003, 119, 2761. (53) Mukamel, S. Principles of Nonlinear Optical Spectroscopy; Oxford: New York, 1995. (54) Kell, A.; Feng, X.; Reppert, M.; Jankowiak, R. On the Shape of the Phonon Spectral Density in Photosynthetic Complexes. J. Phys. Chem. B 2013, 117, 7317−7323. (55) Egorova, D.; Gelin, M. F.; Domcke, W. Time- and FrequencyResolved Fluorescence Spectra of Nonadiabatic Dissipative Systems: What Photons can Tell Us. J. Chem. Phys. 2005, 122, 134504. (56) Hahn, S.; Stock, G. Ultrafast Cis-Trans Photoswitching: A Model Study. J. Chem. Phys. 2002, 116, 1085. (57) Kühl, A.; Domcke, W. Multilevel Redfield Description of the Dissipative Dynamics at Conical Intersections. J. Chem. Phys. 2002, 116, 263. (58) Segal, D.; Nitzan, A.; Davis, W. B.; Wasielewsk, M. R.; Ratner, M. A. Electron Transfer Rates in Bridged Molecular Systems 2. A Steady-State Analysis of Coherent Tunneling and Thermal Transitions. J. Phys. Chem. B 2000, 104, 3817−3829. (59) Kelly, A.; Brackbill, N.; Markland, T. E. Accurate Nonadiabatic Quantum Dynamics on the Cheap: Making the Most of Mean Field Theory with Master Equations. J. Chem. Phys. 2015, 142, 094110. (60) Falke, S. M.; Rozzi, C. A.; Brida, D.; Maiuri, M.; Amato, M.; Sommer, E.; Sio, A. D.; Rubio, A.; Cerullo, G.; Molinari, E.; et al. Coherent Ultrafast Charge Transfer in an Organic Photovoltaic Blend. Science 2014, 344, 1001−1005. (61) Devos, A.; Lannoo, M. Electron-Phonon Coupling for Aromatic Molecular Crystals: Possible Consequences for Their Superconductivity. Phys. Rev. B 1998, 58, 8236. (62) Malagoli, M.; Coropceanu, V.; da Silva Filho, D. A.; Brédas, J. L. A Multimode Analysis of the Gas-Phase Photoelectron Spectra in Oligoacenes. J. Chem. Phys. 2004, 120, 7490. I

DOI: 10.1021/acs.jpcc.5b03989 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C (63) da Silva Filho, D. A.; Kim, E.-G.; Brédas, J.-L. Transport Properties in the Rubrene Crystal: Electronic Coupling and Vibrational Reorganization Energy. Adv. Mater. 2005, 17, 1072−1076. (64) MacKenzie, R. C. I.; Frost, J. M.; Nelson, J. A Numerical Study of Mobility in Thin Films of Fullerene Derivatives. J. Chem. Phys. 2010, 132, 064904. (65) Monahan, N. R.; Williams, K. W.; Kumar, B.; Nuckolls, C.; Zhu, X. Y. Direct Observation of Entropy-Driven Electron-hole Pair Separation at an Organic Semiconductor Interface. Phys. Rev. Lett. 2015, in press. (66) Ishizaki, A.; Fleming, G. R. Unified Treatment of Quantum Coherent and Incoherent Hopping Dynamics in Electronic Energy Transfer: Reduced Hierarchy Equation Approach. J. Chem. Phys. 2009, 130, 234111. (67) Groves, C. Developing Understanding of Organic Photovoltaic Devices: Kinetic Monte Carlo Models of Geminate and NonGeminate Recombination, Charge Transport and Charge Extraction. Energy Environ. Sci. 2013, 6, 3202−3217. (68) Melianas, A.; Pranculis, V.; A. Devižis, V. G.; Inganäs, O.; Kemerink, M. Dispersion-Dominated Photocurrent in Polymer:Fullerene Solar Cells. Adv. Funct. Mater. 2014, 24, 4507−4514. (69) Nelson, J.; Kwiatkowski, J. J.; Kirkpatrick, J.; Frost, J. M. Modeling Charge Transport in Organic Photovoltaic Materials. Acc. Chem. Res. 2009, 42, 1768−1778.

J

DOI: 10.1021/acs.jpcc.5b03989 J. Phys. Chem. C XXXX, XXX, XXX−XXX